Senior Projects Conference

Sepsis is a widespread, potentially life-threatening response in the body to an infection. It is the leading cause of death after a burn injury, causing up to 84% of deaths in adult burn patients. Early detection of sepsis is critical; however, diagnosis in burn patients is more complicated due to the presence of the postburn hypermetabolic response overlapping with typical clinical indicators of sepsis.

To address this, the team, with much help from Dr. Eric Sherer, has used physiological data from the MIMIC-III clinical database to create a predictive computer model that uses logistic regression to analyze a patient’s recorded vital signs and predict whether or not the patient will develop sepsis. From our calculations, breathing rate and heart rate data give the most significant prediction of sepsis onset. The interaction of heart rate and breathing rate (HR+BR+(HR*BR)) reduces statistical deviance the most.

Our model focuses on each patient’s first three hours’ worth of data and averages it to get a truer value and minimize effects of outliers. This will give a prediction of at least 90% accuracy within three hours. This project’s purpose is to create a predictive medical device that will reduce serious infections or fatalities resulting from sepsis in burn patients by providing healthcare professionals with an immediate, accurate prediction of sepsis development.

It is estimated that between 10-20% of Americans will experience a psychogenic nonepileptic seizure sometime in their life, and 2-71% of these will be misdiagnosed as an epileptic seizure. This is due to the fact the epileptic seizures and non-epileptic seizures have similar signs. This can lead to antiepileptic drugs being prescribed to those having a non-epileptic seizures which can have serious side effects. During an epileptic seizure protein prolactin is released by the pituitary gland. Normal human prolactin levels are near zero at 0 to 3 ng/ml in males and nonpregnant females, during an epileptic event, levels can get up 30 ng/ml. Our team is fabricating a handheld device that will assist medical professionals in diagnosing epileptic seizures. For this project, the handheld device must be able to detect a color change between the different levels of prolactin caused by a reaction to prolactin antibodies and horseradish peroxidase.

By performing repeated tests of the color change, we have demonstrated that real-time changes of the prolactin levels can be detected through the color change of the Elisa test strip using a color sensing element. In our preliminary trails, we were able to perform the testing under one hour. This is a huge improvement to the current blood testing method that can take 24 hours. This method of determining prolactin levels shows promise in distinguishing between epileptic seizures and non-epileptic seizures. This can save patients from receiving false diagnoses and allow them to receive the treatment that they truly need.

Babies born prematurely are underdeveloped and need constant medical assistance and supervision because they lack the ability to self-regulate. Current medical devices do not account for the physiological challenges of premature babies, such as their delicate skin or need for direct skin to skin contact with the mother, which aids in the early development of the child. To address current clinical needs the JoeyJack was developed, a nearly wireless monitoring vest that utilizes adhesiveless electrodes and a compact temperature sensor with minimal skin coverage. Our team achieved this through the integration of Maxim MAX30205 temperature sensor, Texas Instrument ADS1293 Analog Front End (AFE) for ECG, adhesiveless fabric-embedded electrodes, and an Arduino Uno. These components were consolidated within a cotton jacket with a separate compact processing unit with a single connecting wire. To determine equivalency of JoeyJacket readings to clinical standards, MATLAB was used to determine signal output variance between the two. The AFE and adhesiveless electrodes produce a signal with a coherence to clinical standard sensors less than 0.75 while reducing child discomfort and abrasions. The JoeyJacket is designed to provide accurate and precise ECG, respiratory rate, and temperature readings while considering the sensitivities of a premature neonate. Although neonates are classified as a rare disease population, the clinical need of the JoeyJacket still exists and will support future generations.

Hundreds of millions of people around the world have some medical condition that puts them at risk of developing ulcers on the sole of the foot. The most significant of these conditions is diabetes mellitus, with roughly 15-20% of people with diabetes experiencing foot ulceration. Approximately 14-24% of foot ulcer cases require limb amputation, with only about a 50% survival rate after the amputation. Normally, a podiatrist looks for early signs of an ulcer with a machine that tests for only either temperature or pressure.

Recent research has shown that the specificity of these machines is low and that multiple measurands are recommended. DuaSock is a wearable smart sock that fulfills this need by measuring two separate factors at once, temperature and pressure, while also allowing the patient to self-monitor their feet from home. This device was designed to predict the development of foot ulcers from early temperature and pressure patterns and warn the patient before the ulcer develops. An array of sensors was implemented along the bottom of the sock to take continuous measurements throughout the day. Trial runs have shown that the sock can reliably measure data in real-life scenarios and determine if an ulcer may be forming.

Parkinson’s disease (PD) is a progressive neurological disorder that results in detrimental gait disruptions,  predisposing PD patients to harmful falls. Because people with PD have fluctuating symptoms, gait assessments made in the clinic do not always reflect the actual disease state of the patient. Our device aims to solve this problem by collecting gait data from PD patients over long time intervals in their native state and providing objective feedback. This device gathers gait data via three Sensoria® IMU sensors: one at the mid-back and one integrated into both socks. We used machine learning to train two algorithms that classify patients into a “high-risk” or “low-risk” category for falling based on gait parameters that are known to change with disease progression (step length and cadence). We also trained an individualized algorithm to differentiate between a person’s normal and impaired gait. Our device will classify the patient’s gait data every 10 seconds and then output the percentage of time the user entered an at-risk gait pattern. Additionally, our device reports the patient’s cadence, stride time variability (STV), and bilateral coordination over time. This device calculates cadence with a percent error of 0.401% and can accurately track STV and report when it drops below the high-risk threshold of 4.4%. This device can serve as a telemedicine tool that provides physicians with objective feedback on the gait status of people with PD.

Total joint replacements are common surgical procedures for people who have a diseased or damaged joint. Older joint implants, originally made in the 90s, often fail, and it is crucial to the patient’s health that the implant is safely replaced. These older implants are held in by bone cement. Bone cement continues to harden over time, which is partially why older implants are particularly difficult to remove. To address this issue, an experiment involving the application of heat to the implant by electrical leads, outputting 44 W, was performed, which weakens the chemical bonds of the bone cement and causes the implant to separate more easily from the bone cement. To test this method, metal bolts were encased in bone cement and select samples were heated up various lengths of time to result in different final temperatures as outer bone cement temperature was a concern for cell necrosis. The force required to remove the bolts from the bone cement was compared between the heated and unheated bolts. The experiment demonstrated that there was a significant difference in the force necessary to pull the bolt out of the bone cement that was heated compared to not heated.

More than 20 million men in America suffer from erectile dysfunction. For many of these men, pills or injections can be used as a treatment for their erectile dysfunction. However, physical conditions, such as prostate cancer, can cause such methods to be ineffective. The only functional solution on the market at this time is the inflatable penile prosthesis (IPP). An IPP involves two inflatable cylinders located in the shaft, a pump in the scrotum, and a saline reservoir placed on top of the bladder. The pump is squeezed to push saline from the reservoir to the cylinders and back again.

However, this current model entails problems both for the user and the surgeon. Problems include difficulty pumping the device, causing it to take more than 30 seconds to inflate, and risk when operating near the bladder. Hydro-Coles accounts for these problems by removing the reservoir and pump and replacing them with an external motorized device. Because of this, the implant inflates in less than 25 seconds and poses less risk to the patient during surgery. In addition, this new IPP requires less time in surgery and is easier to replace if necessary.

The Medtronic FDA approved Laparoscopic Sealer/Divider is a surgical tool primarily used for laparoscopic procedures. The manufacturing process for the device is complex and time-consuming, as it is mostly manually assembled by operators. When defects are found during the production process, particularly after parts have been permanently welded, the materials must be scrapped. Scrapping materials results in a loss of revenue and service for patients in need. In order to reduce manufacturing process defects, the team implemented a Cognex 7000 Series Model 130 Camera to help operators properly align sub-assembly parts prior to entering the laser welder. Using the Cognex software, the system utilizes four customized decision-making tools to notify the operator if the parts are properly aligned prior to passing the part into the laser welder. During the trial run conducted at the Medtronic Boulder, Colorado manufacturing facility, data was collected for 385 devices over a 3-day period. The Cognex failure detection rate was compared to the actual production defect rate. The results indicated that if the Cognex system is implemented permanently by Medtronic, it would notify operators before laser welding, allowing them to have an opportunity to adjust the parts and decrease the total material scrap produced by approximately 30%.